Scholarlink

<p> In the anaerobic fermentation process, the concentration of ammonia nitrogen (NH3-N) in the biogas liquid is an important indicator. This metric can provide a basis for assessing nitrogen nutrition levels in the system, the percentage of various nitrogen types in the substrate, and the decomposition degree of organics. To realize the rapid measurement of NH3-N concentration in biogas liquid during this process, this paper combined the coronavirus herd immunity optimizer (CHIO) algorithm with the idea of interval partial least squares (iPLS) to construct CHIO-iPLS feature interval selection algorithm for selecting high-correlation wavelength intervals of NH3-N for near-infrared spectroscopy (NIRS). We compared its modeling performance with two classical interval selection methods: backward interval partial least squares (BiPLS) and synergy interval partial least squares (SiPLS). Then, the CHIO was combined with BiPLS, SiPLS, and CHIO-iPLS to construct three cascaded wavelength selection methods, donated as BiPLS-CHIO, SiPLS-CHIO, and DCHIO-PLS, respectively. These methods are used for the secondary elimination of redundant wavelength variables within the selected feature intervals. The fusion data of near-infrared transmission spectra collected with different optical pathlength cuvettes were combined with these feature interval selection algorithms and cascade wavelength selection algorithms to construct partial least squares calibration models for NH3-N concentration. Among them, the DCHIO-PLS demonstrated the best modeling performance. The model constructed with 114 feature wavelengths selected by DCHIO-PLS achieved a coefficient of determination of 0.9971, a root mean square error of 10.7663 mg/L, and a mean relative error of 3.9366% for the validation set. This model is capable of meeting the demand to swiftly detect the NH3-N concentration within biogas liquid in the course of anaerobic fermentation. The research results provide a theoretical basis for using NIRS to determine NH3-N concentrations in biogas liquid online.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Rapid detection of ammonia nitrogen concentration in biogas liquid using near-infrared spectroscopy combined with cascade wavelength selection

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月12日

查看全文
  • Weida Wang,
  • Yiyi Wang,
  • Jinming Liu,
  • Chong Tan

摘要

In the anaerobic fermentation process, the concentration of ammonia nitrogen (NH3-N) in the biogas liquid is an important indicator. This metric can provide a basis for assessing nitrogen nutrition levels in the system, the percentage of various nitrogen types in the substrate, and the decomposition degree of organics. To realize the rapid measurement of NH3-N concentration in biogas liquid during this process, this paper combined the coronavirus herd immunity optimizer (CHIO) algorithm with the idea of interval partial least squares (iPLS) to construct CHIO-iPLS feature interval selection algorithm for selecting high-correlation wavelength intervals of NH3-N for near-infrared spectroscopy (NIRS). We compared its modeling performance with two classical interval selection methods: backward interval partial least squares (BiPLS) and synergy interval partial least squares (SiPLS). Then, the CHIO was combined with BiPLS, SiPLS, and CHIO-iPLS to construct three cascaded wavelength selection methods, donated as BiPLS-CHIO, SiPLS-CHIO, and DCHIO-PLS, respectively. These methods are used for the secondary elimination of redundant wavelength variables within the selected feature intervals. The fusion data of near-infrared transmission spectra collected with different optical pathlength cuvettes were combined with these feature interval selection algorithms and cascade wavelength selection algorithms to construct partial least squares calibration models for NH3-N concentration. Among them, the DCHIO-PLS demonstrated the best modeling performance. The model constructed with 114 feature wavelengths selected by DCHIO-PLS achieved a coefficient of determination of 0.9971, a root mean square error of 10.7663 mg/L, and a mean relative error of 3.9366% for the validation set. This model is capable of meeting the demand to swiftly detect the NH3-N concentration within biogas liquid in the course of anaerobic fermentation. The research results provide a theoretical basis for using NIRS to determine NH3-N concentrations in biogas liquid online.